Research Article

Structural basis for blue-green light harvesting and energy dissipation in diatoms

See allHide authors and affiliations

Science  08 Feb 2019:
Vol. 363, Issue 6427, eaav0365
DOI: 10.1126/science.aav0365

All the hues, even the blues

Photosynthetic organisms must balance maximizing productive light absorption and protecting themselves from too much light, which causes damage. Both tasks require pigments—chlorophylls and carotenoids—which absorb light energy and either transfer it to photosystems or disperse it as heat. Wang et al. determined the structure of a fucoxanthin chlorophyll a/c–binding protein (FCP) from a diatom. The structure reveals the arrangement of the specialized photosynthetic pigments in this light-harvesting protein. Fucoxanthin and chlorophyll c absorb the blue-green light that penetrates to deeper water and is not absorbed well by chlorophylls a or b. FCPs are related to the light-harvesting complexes of plants but have more binding sites for carotenoids and fewer for chlorophylls, which may help transfer and disperse light energy.

Science, this issue p. eaav0365

Structured Abstract


Photosynthetic organisms contain light-harvesting antenna systems to gather light energy required for driving photochemical reactions. Diatoms are a group of eukaryotic algae found in fresh water and oceans throughout the world that help form the basis of ocean primary productivity by fixing massive amounts of carbon dioxide into organic carbon. Diatoms are well adapted to this environment in that they contain light-harvesting antennas with exceptional light harvesting and photoprotection capabilities, called fucoxanthin (Fx) and chlorophyll (Chl) a/c-binding proteins (FCPs). FCPs contain the pigments Chl c and Fx, which enable them to absorb light in the blue-green region that is available under water but not effectively used by organisms that contain exclusively Chl a/b. These pigments also confer on FCPs a robust energy-quenching system necessary to thrive in the surface layer of the ocean, an environment with constantly changing light.


FCP proteins belong to the superfamily of transmembrane light-harvesting complex (LHC) proteins with low sequence similarity to the main Lhca (LHCI) and Lhcb (LHCII) subunits of the green lineage organisms. The structures of LHCI and LHCII from higher plants, and the structure of LHCI from a red alga, previously revealed the binding sites for pigments in these antenna proteins. This information was not yet known for FCPs, which limited understanding of the mechanism of light absorption in the blue-green region and energy transfer and dissipation.


We solved the x-ray crystal structure of a dimeric FCP from a pennate diatom Phaeodactylum tricornutum at 1.8-Å resolution. The FCP was purified as a dimer, and the structure showed that two monomers are held together by interactions between their transmembrane C helices. This differs from the predominant organization of trimers found in the major LHCII of the green-lineage organisms. Each FCP monomer binds nine Chls and seven Fxs; the number of Chls is much less than the typical 14 Chls, whereas that of Fxs is greater than the three to four carotenoids found in LHCI and LHCII, resulting in a much higher Fx/Chl ratio in FCP than those in LHCI and LHCII. Among the Chls, two are Chl c located at two sides of the transmembrane helices A and B, and they are in close interaction with two nearby Chls a and one Fx, respectively. This indicates fast energy coupling of Chl c not only with Chl a but also with Fx. Each Fx is surrounded by one or more Chls, suggesting efficient energy transfer between them and also efficient dissipation of excess energy under high light conditions through the abundant Fxs. The binding environment of the two end groups of each Fx showed different hydrophilicities within the protein scaffold, suggesting differences in their preferred absorption region of the blue-green light. One diadinoxanthin (Ddx) molecule is assigned to a position close to the monomer-monomer interface because of its weak electron density, suggesting its easy dissociation from the apoprotein and possible involvement in the Ddx-deepoxidation cycle that functions in energy dissipation.


The FCP structure revealed a network of Chls a/c and Fxs that enables efficient blue-green light harvesting and energy dissipation in diatoms. The ligand structure and binding environment of each pigment revealed in this study will enable detailed studies on the absorption properties of the individual pigments, energy transfer pathways and dynamics, and excess energy dissipation mechanisms in this group of antennas, by both theoretical calculations and time-resolved spectroscopic approaches.

Structure of a FCP.

The FCP proteins are found in thylakoid membranes of diatoms and related algae and function as light-harvesting antennas in these organisms. They bind, in addition to Chl a commonly present in oxygenic photosynthetic organisms, pigments Chl c and Fx, enabling them to harvest blue-green light, which can penetrate into water more efficiently. The structure reveals close interactions between Chls and Fxs, suggesting efficient energy transfer and dissipation among these pigments.


Diatoms are abundant photosynthetic organisms in aquatic environments and contribute 40% of its primary productivity. An important factor that contributes to the success of diatoms is their fucoxanthin chlorophyll a/c-binding proteins (FCPs), which have exceptional light-harvesting and photoprotection capabilities. Here, we report the crystal structure of an FCP from the marine diatom Phaeodactylum tricornutum, which reveals the binding of seven chlorophylls (Chls) a, two Chls c, seven fucoxanthins (Fxs), and probably one diadinoxanthin within the protein scaffold. Efficient energy transfer pathways can be found between Chl a and c, and each Fx is surrounded by Chls, enabling the energy transfer and quenching via Fx highly efficient. The structure provides a basis for elucidating the mechanisms of blue-green light harvesting, energy transfer, and dissipation in diatoms.

Photosynthetic organisms contain light-harvesting proteins in which pigments are positioned to absorb and funnel light to the photosystem core complexes (14) to initiate a series of charge separation and electron transfer reactions. Excess light absorbed by the light-harvesting pigments is known to cause photodamage (1, 57), which may become severe in environments with constantly changing light intensity (8). Several photoprotection strategies have evolved to dissipate excess energy in photosynthetic organisms (1, 3, 811). Diatoms are a group of photosynthetic algae distributed globally in the oceans and fresh water and contribute ~20% of the global primary production (1214). The light-harvesting antenna of diatoms are fucoxanthin-chlorophyll a/c proteins (FCPs), which have exceptional light harvesting and photoprotection capabilities (3, 4, 15) and may contribute to diatoms’ success in environments with exposure to intense and variable light.

Fucoxanthin (Fx) and chlorophyll (Chl) c provide an orange-brown color to diatom FCPs, allowing them to absorb light in the blue-green region (1, 2, 4, 16), which penetrates to deeper water but is not effectively used by the green-lineage photosynthetic organisms (1, 2, 4). Because of the constant circulation of water in the surface layer of ocean, diatoms can experience shifts between weak and strong light in a short period of time (11, 17, 18). Diatoms deal with strong light through a nonphotochemical quenching (NPQ) system, which enables under intense light dissipation of the excess absorbed energy into heat. Diatom FCPs display robust NPQ, which operates under intense illumination (11, 1720) and provides protection from photodamage (2123). The diatom NPQ system is known to involve the diadinoxanthin-diatoxanthin (Ddx-Dtx) xanthophyll cycle (3, 19, 20, 24) manifested by conversion of Ddx to Dtx by the Ddx de-epoxidase under high light and its reverse reaction by the Dtx epoxidase under weak light or dark (1, 3, 20). It is not clear, however, to what extent other factors also contribute to the NPQ in diatoms.

Diatom plastids originated from endosymbiosis of an ancestral red alga (13), and their FCPs belong to the superfamily of light-harvesting complex (LHC) proteins with low sequence similarities to the main Lhca (LHCI) and Lhcb (LHCII) proteins of the green lineage organisms (2, 25). The major subunits of diatom FCPs are encoded by lhcf genes, which have more than 10 genes in each diatom species (26, 27). The energy harvesting and dissipation features of FCPs are remarkably different from those of LHCs found in red algae and green plants (5, 1216) because of the binding of different Chls and carotenoids (Chl c and Fxs) and likely also due to different arrangements of the pigments. The sequences of lhcf genes are highly similar, and the FCP proteins produced from these genes are also considered to have similar structure and properties, including binding of similar numbers of pigments (2, 22, 28). Biochemical and biophysical studies have been conducted on the light harvesting, energy transfer, and dissipation functions of FCPs (3, 11, 1525, 29). However, the arrangement of the pigments involved in these reactions is not currently known, which limits our understanding of these processes at a molecular level.


Overall structure of the FCP

Here, we report the x-ray crystal structure of an FCP from a widely distributed, marine pennate diatom Phaeodactylum tricornutum at 1.8-Å resolution (Fig. 1, figs. S1 and S2, and table S1). This FCP was isolated in a homodimeric form (Fig. 1 and fig. S3) and corresponds to the product of lhcf3 or lhcf4 genes registered in the National Center for Biotechnology Information (NCBI) database (these two genes code for identical sequences), according to the results of N-terminal sequencing and mass spectrometric analyses. Phylogenetic analysis showed that the Lhcf4 protein is similar to both LHCI and LHCII proteins, with several conserved residues for the binding of Chls (fig. S4A), and higher similarities were found among the different Lhcf proteins, especially among Lhcf1 to Lhcf12 (fig. S4B). We confirmed that the dimeric form of the isolated FCP is present in solution (fig. S3, A, B, and C). The protein-pigment complex has an apparent molecular weight of 64 kDa estimated from size exclusion chromatography, which is consistent with sum weight of the apoproteins (~37 kDa per dimer) and pigments (Chls and Fxs, ~27 kDa) for a dimer. The isolated complex is much smaller than the LHCII trimer (~145 kDa) isolated from the green alga Bryopsis corticulans under similar conditions (fig. S3C).

Fig. 1 Overall structure of FCP from P. tricornutum at a resolution of 1.8 Å.

(A) Overall structure of the FCP dimer with a view from a direction parallel to the membrane plane. The dashed line in the center shows symmetric axis of the two monomers. Protein structure in the left-side monomer is depicted in cyan, whereas that in the right-side monomer is depicted in gray. Letters in red shows the transmembrane helices. Color codes for cofactors are Chl a, green; Chl c, magenta; Fx, orange; Ddx, marine; phosphatidyl-glycerol, violet; digalactosyl-diacylglycerol, limon; OTG, pale cyan; α-DDM, light blue. (B) Arrangement of the pigments (Chls and Fxs) in the FCP dimer with the same view as in (A). (C) The structure of an FCP monomer, including water molecules (light pink) and two calcium ions (light blue) located in the stromal side. (D) Monomer-monomer interactions between the two C helices with a view from the stromal side. Hydrogen bonds are depicted in dashed lines in magenta, whereas hydrophobic interactions are depicted in yellow dashed lines.

Within the crystal structure, two FCP monomers meet at a dimer interface between the C-helices (Fig. 1). Two Chl a406 molecules insert a part of their tetrapyrrole rings into the gap space of the two helices C to form hydrophobic interactions, and the phytol tails of the two Chl a molecules also interact with two Ddx molecules to keep them at the position close to the monomer-monomer interface (Fig. 1D). In addition, Arg104 of each monomer is hydrogen-bonded with Ser100 of the adjacent monomer (Fig. 1D and fig. S5, E and F) at the stromal surface. These interactions thus held the two monomers to form a dimer.

Each FCP monomer contains seven Chls a, two Chls c, seven Fxs, one Ddx, two calcium cations, one phosphatidyl-glycerol, and one digalactosyl-diacylglycerol. In addition, four n-octal-β-d-thioglucoside (OTG) and one n-dodecyl-α-d-maltopyranoside (α-DDM) used for solubilization and crystallization were found at the lumenal side and close to Chls and carotenoids, suggesting their possible stabilization roles for the FCP during purification and crystallization.

The secondary structure of each monomer is similar to those of Lhca (or Lhcr) and Lhcb proteins (Fig. 2A and figs. S4A and S5) (3035); however, the N- and C-termini and loop regions are shorter in the FCP. This is reflected by the large differences in the root mean square deviations between the Cα atoms of Lhcf4 (FCP) and Lhcb1, which range from 0.7 to 1.2 Å for the three transmembrane helices but from 9.8 to 11.0 Å for the N- and C-termini and loop regions. In the N terminus, a large loop region extended from the beginning of helix A in both Lhca and Lhcb at the stromal surface was changed to a much shorter loop in Lhcf4, whereas in the C terminus, helix D found at the lumenal surface was absent in Lhcf4 (Fig. 2A and figs. S4A and S5). As a result, the characteristic N-terminal “WYGPDR” motif and a C-terminal Trp residue crucial for the LHCII trimer formation (3133) are absent in this Lhcf4 (fig. S4A), which may account for why Lhcf4 forms a homodimer and no trimer is found in the present study. In addition, helix C of Lhcf4 is more tilted against the membrane plane in part to facilitate the formation of the dimer through hydrophobic interactions between them, making it shifted relative to its counterpart in Lhca and Lhcb (fig. S5D). Hydrogen bonds and salt bridges contribute to the structures of the loop regions in both stromal and lumenal sides and around the calcium ion in the lumenal side (fig. S6). Acidic residues located at the lumenal surface (fig. S6) may be involved in pH-induced conformational changes during energy dissipation (810, 36, 37).

Fig. 2 Comparison of the structures between FCP and LHCII.

(A) Superposition of the structures of an FCP monomer (Lhcf4, red) with LHCII [Lhcb1, Protein Data Bank (PDB) code 1RWT, marine] (31). The apoproteins are depicted in ribbon, and all of the cofactors are removed for clarity. The areas encircled by dashed lines indicate structures with large differences. The helices are labeled with red letters. (B) Superposition of the Chls of FCP with those of LHCII. Chls a and Chls c of FCP are depicted in green and magenta, and Chls a, b of LHCII are depicted in blue and light gray, respectively. (C) Superposition of the carotenoids of FCP with those of LHCII. Fxs are depicted in orange, Ddx in marine, lutein in yellow, vioxanthin in lime, and neoxanthin in purple.

Arrangement of Chls and energy transfer

FCP contains two Chls c in addition to seven Chls a (Figs. 1, A to C; 2B; and 3 and table S2); the Chl a/c ratio of 3.5 is close to the ratio of 4 as determined with biochemical analyses for the samples used in this study (fig. S7). The total number of Chls in an FCP monomer, nine, is far lower than the typical 14 found in most LHCII and LHCI monomers (table S2). Four Chls a (a402, a404, a406, and a407) and the two Chls c are found in similar positions within the structure of LHCII (and LHCI) (Fig. 2B and fig. S8). The other three Chls a have different positions or orientations with those in LHCII, among which, Chls a405 and Chl a409 are found at the lumenal side close to helices C and A, and Chl a401 is present in the N-terminal loop region (Figs. 2B and 4C). Eight Chls in FCP are coordinated by two His, three Glu, and three Gln residues, whereas Chl a401 is coordinated by a water molecule (fig. S6D and table S3). This coordination pattern is largely different from those of the inner antennae of PSII and PSI core (34, 35, 38, 39) in which most of the Chls are coordinated by His residues. The larger variation of Chl ligands in FCP (and LHCI and LHCII) than that in PSI and PSII core antennae may reflect the evolution of peripheral antenna proteins to cope with differences in environment. The absence of Chl b in FCP may therefore be a result of independent diversification of FCP from LHCII in the green lineage to accommodate binding sites for the different pigments such as Chl c and Fx, which extended its capacity to absorb different qualities of light (3, 11, 15).

Fig. 3 Binding environments of two Chl c molecules.

(A) Overall position and coordinating environments of the two Chl c molecules. Coordinating bonds for the two Mg2+ ions are depicted in solid lines, whereas hydrogen bonds are depicted in dashed lines. Color codes used are the same as in Fig. 1. (B and C) Interactions of Chl c408 (B) and Chl c403 (C) with nearby Chl a and Fx molecules. Distances for hydrogen bonds and π-π interactions are given in angstroms.

Fig. 4 Distribution of Chls in FCP and possible energy transfer pathways among them.

(A) Distribution of the Chls with a view along the membrane plane. The adjacent Chls in the same layers are connected with dashed lines, and their center-to-center distances are labeled (angstroms). (B) The same view as in (A), with the distances between the stromal and lumenal side layers indicated. (C and D) Distribution of Chls with views from the stromal side (C) and lumenal side (D). Color codes for the pigments are the same as those in Fig. 1.

The two Chls c are located in the two sides of the two crossing transmembrane helices A and B and are coordinated by His39 (Chl c403) and Gln143 (Chl c408), respectively (Fig. 3). The two polar C-17 propionic acids of these Chls c interact with alkaline residues Arg31 and Lys136, respectively, through ionic bonds, whereas their C-173 carboxyl oxygen is strongly hydrogen-bonded to the hydroxyl group of the cyclohexane (C) end groups of Fx306 and Fx307 (Fig. 3, B and C, and table S3). The presence of these hydrophilic groups surrounding the tetrapyrrole rings of the two Chls c is in agreement with a higher hydrophilicity of Chl c than Chl a owing to the lack of the long phytol tail (40). The close relationship of Fx306 and Fx307 with the two Chls c imposes restraints on the species of Chls that can occupy these positions because the phytol tails of Chl a will apparently be in conflict with the two Fxs. Because carotenoids corresponding to Fx306 and Fx307 are absent in LHCI and LHCII (Fig. 2C), this may account for why these positions bind Chl a in the green lineage but Chl c in FCPs. In addition to the hydrophilic groups, each of the two Chl c molecules are in close interactions with two Chl a molecules (Fig. 3, B and C).

The Chls in FCP are distributed in layers on the lumenal and stromal sides of the thylakoid membrane (Fig. 4). In the stromal layer, six Chls (four Chls a and two Chls c) form two coupled Chl a-c-a clusters (a402-c403-a406, and a401-c408-a407) in each monomer (Fig. 4, A and C), among which, Chls c403 and c408 in each cluster are strongly coupled with Chls a406 and a401, with partial overlaps of their tetrapyrrole rings and a closest π-π distance of 3.4 and 3.9 Å, respectively (Fig. 3, B and C, and table S4). On the other hand, the π-π distances between the tetrapyrrole rings of Chls c403-a402 and Chls c408-a407 are at 5.6 and 6.1 Å, respectively (table S4). These close π-π distances enable fast and efficient excitation energy transfer between Chl c and neighboring Chl a at two directions within the two Chl a-c-a clusters, explaining the speed at 60 to 100 fs and the efficiency at 100% of energy transfer between Chl a and Chl c previously determined with time-resolved spectroscopy (15, 40, 41). The π-π distance between the two clusters is 9.6 Å within the same FCP monomer (Chl a401-Chl a402), but 6.4 Å with the adjacent monomer (Chl a406-a406*) (table S4), suggesting fast and efficient energy transfer between the two monomers in the stromal layer and an advantage for the formation of the dimer. These structural features, in combination with the intrinsic characteristics of the pigments, make Chl c an efficient harvester of blue-green and even yellow light, which is the “green gap” where Chls a and b absorb weakly. The energy absorbed by Chls c is then transferred efficiently to the coupled Chls a (1, 4).

The remaining three Chls (a404, a405, and a409) are located close to each end of the three transmembrane helices at the lumenal side (Figs. 2B and 4, B and D), with an edge-to-edge distance of ~8.4 Å between Chls a404 and a405 and a much longer distance between Chls a404 and a409 (table S4). The π-π distance between the stromal and lumenal layers is 8.9 Å (Chl a404-a406) and 11.8 Å (Chl a409-a402) in a monomer (Fig. 4B), suggesting a slower excitation energy transfer between the two layers, which is similar to what has been reported for LHCII and LHCI (3135). Because the seven Chl a molecules in FCP are clearly separated, strong Chl a-c coupling will be dominant, whereas Chl a-a coupling will be rather weak based on the present structure. This organization of the Chl a molecules suggests similarities in their spectral properties, which is in agreement with previous spectroscopic results showing that all of the Chls a in FCP have similar energetic levels (15, 28). Therefore, it would be difficult to determine the trap and exit sites of the absorbed excitation energy on the basis of the excitonic level of individual Chls. However, on the basis of the current structure, we propose that the lateral Chl a401 strongly coupled with Chl c may be responsible for collecting absorbed excitation energy and transmitting it to the reaction center or adjacent antennas.

Arrangement of FXs and blue-green light absorption, energy dissipation

FCP binds seven Fxs and one Ddx (Figs. 2C and 5A). By contrast, each of the LHCI and LHCII subunits bind only three or four carotenoids (Fig. 2C and fig. S8) (3035). Fx303 and Fx305 bind to the conserved lutein sites in LHCII, with their polyene chains embedded in the grooves formed by the crossing-helices A and B, and their end groups are more twisted, whereas the remaining five Fxs and the only Ddx bind in positions in which no corresponding carotenoids are found in LHCII (Fig. 2C and fig. S8). The polyene backbones of six Fxs are inclined inside the hydrophobic membrane, and their epoxycyclohexane (E) and cyclohexane (C) end rings are oriented toward the membrane surfaces (Fig. 5A and table S5). The remaining Fx304 is located at the stromal surface, with an angle almost horizontal to the membrane under the “tongue” region of the C-A loop, whose polyene chain is shielded by side chains of Ile121 and Phe123 at the stromal side. This carotenoid binding mode has not been observed in LHC proteins (3035, 42), and we suggest that it may be involved in tuning pigment wavelength.

Fig. 5 Distribution of Fxs and Ddx and the possible quenching site in the FCP dimer.

(A) Side view of the distribution of Fxs with the protein subunits depicted in a surface model. (B to G) Interactions of the individual Fxs with nearby Chls, with their closest distances for the π-π interactions given in angstroms.

Fx has large solvent effects similar to that seen in peridinin and siphonaxanthin (4346): An increase in the polarity of the solvent causes a large redshift in their absorption due to the presence of the conjugated carbonyl groups in these carotenoids. The varied binding environments of Fxs in FCP imply different spectroscopic properties. Because Fx303 and Fx305 are located in the central region of the membrane, they may absorb light predominately in the “blue” region because of their hydrophobic binding environment. Fx301 and Fx302 are suggested as “green” molecules for the presence of polar pockets surrounding their C-end groups, whereas Fx306 and Fx307 are proposed as “red” molecules for their polar binding environments of both end groups to make their absorption largely red-shifted to 500 to 550 nm region (29, 4548). In fact, although the E-rings of Fx306 and Fx307 are located at the lumenal surface, their C-rings are located inside the membrane. However, these two Fxs are close to the two Chl c molecules, with their polyene chains running parallel to the tetrapyrrole rings of the two Chls c at ~3.5 Å and their C-rings hydrogen-bonded to the two propenoic acids of the two Chls c molecules (table S5) as shown above (Fig. 3, B and C, and table S5), making the environment of the C-rings of these two Fxs also rather hydrophilic. Similarly, the distinct Fx304 may also be a “red Fx” because its two end groups are located at the stromal surface close to the bulk solution and also to the Chl c molecules. These structural features allow Fx to play the major light-harvesting role, particularly under water, where light in the red region useful by Chl a diminishes quickly and more “green” photons arrive (3, 7).

The remarkably high Fx/Chl ratio in FCP effectively results in every Fx being surrounded by one or more Chls (Figs. 3, B and C, and 5, B to G), where π-π interactions between Fxs and the tetrapyrrole head groups of Chls are found at distances of ~4.0 Å in most cases (table S4). These close contacts enable fast and efficient excitation energy transfer between Fxs and Chls, which has been shown to be as fast as 75 fs and with an efficiency beyond 90% for the transfer of absorbed green light from Fx to Chl a (15, 41, 48), ensuring Fx to efficiently harvest and use photon energy in the “green-gap” region. However, energy transfer from Fx to Chl c in the major FCP complexes has not been observed so far (48). Our structure suggested that this kind of energy transfer should be very fast and therefore may have escaped from detection. On the basis of the current structure, the energy harvested by Fx306 and Fx307 must be transferred to Chl c403/Chl c408 located between Fx306/Fx307 and Chl a406/Chl a401 (Fig. 3, B and C) before being transferred to the Chl a molecules. Thus, time-resolved spectroscopic studies with a higher time resolution are required to clarify the possible energy transfer from Fx to Chl c.

The close association of every Fx with respective Chls also suggests that the excess energy absorbed by Chls may be dissipated quickly and efficiently through their nearby Fxs, thus protecting the photosynthetic apparatus from photodamage (3). Excessive irradiance will produce triplet Chls and reactive oxygen species at the photosynthetic reaction centers (37), causing photodamage to the photosynthetic apparatus. Photoprotection is a mechanism by which the excess energy is dissipated as heat, protecting photosystems from damage (3, 37). Because diatoms live under rapidly fluctuating light environment owing to the rapid mixing in the surface of ocean (3, 7), they have gained a large amount of light-harvesting antenna proteins (FCPs) to harvest energy effectively under low light conditions. However, when diatoms are brought to the surface layer of ocean in a short time (3), the large amount of FCPs will absorb too much light, which may cause photodamage. Under such conditions, photoprotection is triggered by the Ddx-Dtx de-epoxidation reaction, transmembrane proton gradient, and/or reorganization and conformational changes of the FCP complexes (8, 37). In the present structure, we assigned one Ddx in the vicinity of the twinned helices C in the interface of the two monomers (Figs. 1 and 5). The electron density for the Ddx molecule is weak (fig. S2C), suggesting that it may be disordered or present at low occupancy because of weak interactions with protein or other ligands. This is in agreement with the biochemical analysis of our purified sample showing that there is less than one Ddx per FCP monomer (fig. S7) and may explain why the amount of Ddx obtained from biochemical analyses varies widely in the literature (fig. S7) (15, 16, 28, 49, 50) because Ddx may be easily lost during purification. The apparent weak binding of Ddx may facilitate exchange through the Ddx-Dtx cycle.

Acidification of the lumenal space has been shown to trigger reorganization and/or conformational changes of the antenna proteins, initiating another mechanism of energy dissipation (8, 37). Several acidic residues (Glu54, Asp64, Glu72, Glu82, and Glu158) were found to be exposed to the lumenal surface in the FCP structure (fig. S6A); these residues may be related with the protonation-induced conformational changes that function in photoprotection. Among these residues, Glu72 and Glu82 are coordinated to two calcium cations at the lumenal surface (fig. S6B), suggesting that pH-induced reorganization (8, 51, 52) may be modulated by divalent cations. In addition, conformational changes of Fx301 and Fx302 may be triggered by protonation of acidic residues in the BC- and/or C-terminal loops, leading to the quenching of excitation energy from the nearby Chls similar to that seen with the central lutein in LHCII (810, 36, 37). Furthermore, Fx301 is physically coupled with Chls a401 and a409 (Fig. 5) and therefore also prone to undergo conformational changes upon aggregation of FCP, as seen for the corresponding neoxanthin in LHCII (8, 11, 15, 37). These structural features indicate that the arrangement of Fxs and Chls in the FCP dimers is optimized to assure the efficient light-harvesting in the blue-green region as well as energy dissipation under a highly fluctuating light environment.


The structure of FCP reveals an arrangement of Chls a/c and Fxs, ensuring efficient blue-green light harvesting and energy transfer and dissipation in diatoms, a dominant phytoplankton. We anticipate that the structural information provided by this study will greatly promote theoretical and time-resolved spectroscopic studies to elucidate the energy migration and dissipation mechanisms in this group of antenna proteins.

Materials and methods

Cell culture and FCP purification

Cells of the pennate diatom Phaeodactylum tricornutum (FACHB-863, Freshwater Algae Culture Collection at the Institute of Hydrobiology, Wuhan, China) were grown in a sterile artificial seawater F/2 medium at 22°C under continuous light (40 μmol photons m−2 s−1) for one week. The cells were harvested and resuspended in an ice-cooled medium containing 20 mM tricine, 10 mM MgCl2, 20 mM KCl, 5% sucrose, pH 7.8 (TMKS buffer), and then disrupted by glass beads. Thylakoid membranes were collected by centrifugation at 100,000 × g for 20 min at 4°C, followed by solubilization with 1% (w/v) n-dodecyl-α-D-maltopyranoside (α-DDM) and at a concentration of 0.5 mg Chl a/ml for 30 min on ice based on the method used for isolation of LHCII from a green alga Bryopsis (44) and Cyclotella (53). After centrifugation to remove unsolubilized membranes, the FCP enriched supernatant (FCP pool) was loaded onto a Q-Sepharose HP column (GE Healthcare) pre-equilibrated with the TMKS buffer containing 0.03% (α-DDM). The column was washed with 0.25 M NaCl to remove most of contaminating proteins, followed by elution with a linear gradient of 0.25 to 0.42 M NaCl. The Lhcf4-rich fraction (labeled Lhcf4) was eluted at 0.35-0.38 M NaCl (fig. S3A), which was collected, concentrated with an AMICON centriprep-50 (cut-off molecular weight: 50 kDa) filter and then purified with a second Q-Sepharose HP column with the same conditions as the first column, to increase the purity of the FCP. The eluted FCP was concentrated again and loaded onto a linear sucrose gradient with 5 to 20% sucrose in the TMK buffer containing 0.03% α-DDM, followed by centrifugation at 303,800 × g for 16 hours (fig. S3B). Two bands were obtained, among which, the upper band is an FCP monomer and the lower band is an FCP dimer based on their apparent molecular masses analyzed by gel filtration chromatography (fig. S3C, the elution profile for the FCP monomer was not shown for clarity). The band containing the FCP dimer was collected and concentrated by precipitation with 25% PEG1000, and used for crystallization. The oligomerization states of the “FCP pool” solubilized from the thylakoid membranes by the detergent and the purified FCP were analyzed by gel filtration with a Superdex 200 PC3.2/30 column (2.4 ml, GE Healthcare) (fig. S3C). For comparison, dissolved FCP crystals and LHCII trimers from a green alga Bryopsis were also analyzed by the gel filtration (fig. S3C). The results showed that the FCP pool before separation by the ion-exchange chromatography contained mainly dimers and monomers, whereas the purified FCP and re-dissolved crystals had a molecular mass equivalent to an FCP dimer.


Crystallization trials were performed using the oil batch method at 293 K in which, 3.0 μL of the FCP-dimer dissolved in 10 mM MES [2-(N-morpholino)ethanesulfonic acid; pH 6.5] at 2.0 mg Chls ml−1 (Chls a and c) was mixed with an equal volume of a precipitate solution containing 100 mM MES (pH 6.5) or Tris [tris(hydroxymethyl)aminomethane; pH 8.5], 22 to 26% PEG 1000, 100 mM CaCl2, 3% (w/v) n-octal-β-D-thioglucoside (OTG, Anagrade, Anatrace), and then covered with 15 μL paraffin oil (Hampton). Crystals appeared within 2 to 3 days and reached a maximum size of 0.20 mm × 0.10 mm × 0.05 mm within 5 days, and exhibited a dark-orange color and elongated rectangular shape (fig. S1, A and B). We confirmed that the FCP dimer in the crystal retained the similar absorption and fluorescence properties before crystallization (fig. S9). For the x-ray diffraction experiments, the crystals were transferred into a cryo-protectant solution containing 45% PEG 1000, 50 mM MES (pH 6.5) or Tris (pH 8.5), 50 mM CaCl2, 1.5% (w/v) OTG and incubated for 5 min, and then flash-frozen in a nitrogen stream at 100 K.

Data collection, processing, structural determination, and refinement

Screening of the resolutions of crystals and native data collection were performed at beamlines BL17U1 of Shanghai Synchrotron Radiation Facility, China (54) and BL41XU of SPring-8, Japan. Several thousands of crystals were screened, among which the best crystals diffracted to a resolution of 1.63 Å (fig. S1 and table S1). Diffraction images were recorded at an X-ray wavelength of 1.0 Å, with an exposure time of 0.1 s and detectors of ADSC Q315r CCD for BL17U1 and Pilatus 6M for BL41XU. The data set was collected by rotating the crystal with a 0.2° oscillation angle over a range of 130°. The FCP crystals obtained belonged to the space group P212121 (grown in the pH 8.5 Tris buffer) or C2221 (grown in the pH 6.5 MES buffer), with unit cell dimensions a = 47.79 Å, b = 123.31 Å, c = 140.71 Å, and a = 47.75 Å, b = 115.72 Å, c = 141.26 Å, respectively (table S1). Both types of the crystals showed a severe anisotropy in the diffraction intensities in the axis b, so ellipsoidal truncation was performed with the STARANISO server ( (55). The processing results from this server suggested an overall resolution limit of 1.64 Å with a CC1/2 of 0.3, as well as resolutions of 1.64 Å and 1.79 Å, respectively, based on a criteria of mean I/sigma (I) over 1.5 or 2.00. Due to the high anisotropy of the diffraction data, the resolutions of different directions were suggested to be 1.64 Å, 2.82 Å and 1.72 Å, respectively, for the a, b and c axes with a criterion of mean I/sigma(I) over 1.50. Structural refinement performed at the later stages showed that gap between Rfree and Rwork became larger and the Rfree was also increased when the resolution was set to higher than 1.80 Å. Thus, we determined the overall resolution of our structure to be 1.80 Å.

Single-wavelength anomalous diffraction (SAD) data sets were collected from a single crystal at beamline X06DA of the Swiss Light Source at the Paul Scherrer Institute (Switzerland) at an X-ray wavelength of 2.075 Å, using a multi axis goniometer, PRIGO and a Pilatus 2M dector (56). Due to the high anisotropy and low isomorphism of the crystals, a very high redundancy of the native SAD data had to be collected from a single crystal in order to obtain the phase information. All diffraction data was processed, integrated and scaled using the XDS Program Package (57).

The native SAD data set was indexed and scaled to 2.7 Å resolution, which showed a redundancy of 123 for the whole resolution range and 110 for the highest-resolution shell, respectively. Based on the SAD data, phase information was calculated with Crank2 in CCP4i2 (56, 58), which gave rise to 20 apparent anomalous peaks in the anomalous difference Patterson map for a dimer. These peaks were later identified as 4 calcium atoms, 8 sulfur atoms from methionine, 7 magnesium atoms from Chls and 1 sulfur atom from OTG in each asymmetric unit. Iterative substructure improvement and phasing were performed with REFMAC5 and PEAKMAX, and the handedness of substructures was determined by Solomon in Crank2 of CCP4i2. Density modification was performed with Parrot with Fourier recycling, and BUCCANEER was used for combined iterative model building with density modification and phase refinement in Crank2 of CCP4i2. Pigments, lipids and cofactors were incorporated and the amino acid residues were manually modified using COOT (59). The phase was then extended to the final resolution of 1.8 Å with the native dataset using molecular replacement by PHASER (60). Residues 1 to 166 out of the total 167 residues were traced in the model. The quality of the structure was analyzed using Procheck (61). All figures were prepared using Pymol (62).

Assignment of the light harvesting cofactors

Chls a and c were identified based on the existence of a phytol chain for Chl a, and the planarity of C-181, C-18, C-17, and C-171 given by the C-18=C-17 double bound for Chl c. While most of the Chl a have clearly identifiable phytol tails, Chl a405 has almost no clear electron density for the phytol tail and its head group may be better fitted with a Chl c; however, based on the pigment analysis it was assigned as a Chl a. The two Chl c sites were further differentiated into Chl c1 and Chl c2, and structural refinement with a loose restraint in the distance between the ethyl and ethylene groups of C-8 provided the tentative assignment of Chl c403 as Chl c2 and Chl c408 as Chl c1, respectively. The electron density for the Ddx molecule was also rather weak especially in its head group, and this site could be occupied with a lower occupancy (or even by other carotenoids). The orientation of the Ddx molecule is assigned with its epoxy group directed toward the stromal side in the present structure. When we reverse the orientation of this molecule and performed the structural refinement again, we found that the Rfree value of the whole structure increased by 0.2%, and the B-factor of the Ddx molecule increased from 87 to 90. Thus, we consider that the current orientation of the Ddx molecule is consistent with our electron density map.

Pigment analysis

Pigments were extracted from the FCP sample with 90% (v/v) acetone and the Chl concentration was calculated as previously described (63). HPLC was performed with a C-18 reversed-phase column (ϕ 4.6 mm, length 250 mm, 5 μm particle size, Grace, USA) in a Waters e2695 separation module equipped with a Waters 2998 photodiode array detector. The pigments were eluted at a flow rate of 1 ml/min using a 20 min linear gradient with 0 to 100% solvent B (ethyl acetate) mixed with solvent A (methanol:water = 90:10), which was then continued isocratically with 100% solvent B for 2 min. Pigments were detected by their absorbance at 445 nm.

The authentic standard of Chl a was purchased from Sigma. For commercially unavailable standards (Chl c, Fx. and Ddx), the pigments were extracted from the P. tricornutum thylakoid membranes. The extinction coefficients for Chl c (c2 type) and other pigments used were taken from refs. 5 and 36. Standard curves for the quantification of pigments were obtained by plotting the amount of the pigments loaded onto the HPLC column against the area of each elution peak, with each data point obtained from the average of three independent measurements. For comparisons, other values reported previously for different kinds of samples were summarized in fig. S7 (15, 16, 28, 49, 50).

Peptide analyses and spectrophotometric measurements

Protein composition of the FCP samples was analyzed using SDS-PAGE with a gel containing 16% polyacrylamide and 7.5 M urea. For identification of the FCP subunit, N-terminal sequencing, mass spectrometric analyses and spectrophotometric measurements were performed as described previously (44, 64). The N-terminal sequence obtained was AFEDELGAQPPLGFF, which was used together with several internal sequences from digested peptides to search homologous sequences in the NCBI databases, which gave rise to two sequences encoded by lhcf3 and lhcf4 of P. tricornutum, respectively. These two genes encoded exactly the same sequence with a total of 167 amino acid residues. Thus, the FCP subunit was identified to correspond to the product of lhcf3 or lhcf4. Alignment of homologous sequences acquired from NCBI database and phylogenetic analysis were performed with tools provided by the website (65).

Absorption spectra were measured at room temperature with an ultraviolet-visible spectrophotometer (UV–Vis 2550, Shimadzu, Japan) at 10 μg Chl ml−1 in the TMK buffer containing 0.03% α-DDM. Low-temperature (77 K) fluorescence emission and excitation spectra were measured using a fluorescence spectrophotometer (F-4700, Hitachi, Japan) at a Chl concentration of 2 μg Chl ml−1 for FCP dimer in the TMK buffer supplemented with 30% glycerol and 0.03% α-DDM. The spectral sensitivity of the fluorescence spectrophotometer was corrected using a light source with a known radiation profile (Hitachi, Japan).

Supplementary Materials

Figs. S1 to S9

Tables S1 to S5

Reference (66)

References and Notes

Acknowledgments: We thank M. Sang and D. Chen for their help with the culture of the diatom cells, N. Matsugaki for help on the collection of the SAD data, the CCP4 workshop team (SPring-8 Japan, 2017) for advices on analyzing the native-SAD data and phasing, and H. Lin for discussions. Diffraction data at 1.0-Å wavelength was collected at beamlines BL17U1 of Shanghai Synchrotron Radiation Facility (SSRF, China) and BL41XU of SPring-8 (Japan); Native SAD data was collected at beamlines BL1A of Photon Factory (PF, Japan) and X06DA of Swiss Light Source at the Paul Scherrer Institute, and we thank the staff members of these beamlines for their extensive support. Funding: This work was supported by the National Key R&D Program of China (2017YFA0503700), a Strategic Priority Research Program of CAS (XDB17000000), a CAS Key Research program for Frontier Science (QYZDY-SSW-SMC003), a National Basic Research Program of China (2015CB150100 to T.K.), and JSPS KAKENHI no. JP17H0643419 of MEXT, Japan (to J.-R.S.). Author contributions: J.-R.S., W.W., and T.K. conceived the project; C.X., S.Z, and W.W. cultured the cells and purified the FCP complex; W.W. and C.X. grew and optimized the FCP crystals; W.W., C.X. and L.-J.Y. conducted the diffraction experiments and collected the data at 1.0 Å wavelength; W.W., L.-J.Y., Y.U., and T.T. collected the diffraction data for SAD phasing; W.W., and L.-J.Y. determined the phase; L.-J.Y., M.S., and W.W. refined the structures; W.W., L.-J.Y., T.K. and J.-R.S. wrote the manuscript; and all authors discussed and commented on the results and the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Atomic coordinates have been deposited in the PDB under the accession no. 6A2W, in which four new ligands were assigned as A86 (Fx), DD6 (Ddx), KC1 (Chl c1), and KC2 (Chl c2), respectively. All other data are presented in the main text or supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article